IPv4 vs IPv6 on Automatic Configuration

In world of IPv4, DHCPv4 provides a perfect way for IPv4 automatic configuration including IP address, DNS and routing entires. In world of IPv6, above works has been split into types:

The IPv6-RA is designed to provides fundamental network configuration to host with minimum cost. The DHCPv6 is designed to provides additional and extensive network configuration.

Since DHCPv6 does not provides routing configuration, IPv6-RA is mandatory in order to serve the automatic IPv6 configuration.

The brief difference between IPv6-RA and DHCPv6 are:

    |            |  IPv6-RA |   DHCPv6  |
    +------------+----------+-----------+
    | IP Address |  /64     | /128      |
    | Routing    | Yes      | No        |
    | DNS        | Yes      | Yes       |
    | NTP        | No       | Yes       |

  • IP Address configuration

    • IPv6-RA: Host use SLAAC to generate IP address from router provided /64 prefix.
    • DHCPv6: DHCPv6 server provides hosts a /128 address which could be used to create the DNS AAAA entry for the acknowledged host.

  • DNS Configuration

    • IPv6-RA: Host use IPv6-RA DNS RA options to get DNS configuration from route.
    • DHCPv6: DHCP server provides the DNS configuration to hosts.

  • Routing

    • IPv6-RA: Host use Router Advertisements
    • DHCPv6: Not standardized, hence DHCPv6 should not provide routing information.

  • NTP:

    • IPv6-RA: Do not provide NTP time server configuration.
    • DHCPv6: DHCP server provides the NTP configuration to hosts.

Server side configuration: dnsmasq

The dnsmasq project provides IPv6-RA and DHCPv6 with simple configuration for common use cases.

For more detail, please refer to manpage of dnsmasq. A quick example would be:

1
2
3
4

# Enable IPv6-RA
enable-ra
# Use the IPv6 prefix on eth1 interaface, and 0x1 to 0xff as DHCPv6 range.
dhcp-range=tag:eth1,::1,::ff,constructor:eth1,ra-names,64,48h

Above configuration are using ra-names method for DHCP range, as a result, the host will get:

  • A /64 address generated using SLAAC algorithm. Ignore the DHCPv6 range setting.
  • A /128 address chose by DHCPv6 server from above DHCPv6 range. And this address if pingable, will be the DNS AAAA entry for the host.
  • The DNS options and defaults routes get from IPv6-RA. Host might use DHCPv6 for DNS option if DNS RA options is not supported.

Example output on the host(The DHCPv6 server has the 2001:db8:1::1 address):

[fge@fed ~]$ ip -6 addr show dev  ens10
3: ens10: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1280 state UP qlen 1000
    inet6 2001:db8:1::1f/128 scope global dynamic noprefixroute
       valid_lft 6932sec preferred_lft 6632sec
    inet6 2001:db8:1:0:415b:bc93:c47b:76c8/64 scope global dynamic noprefixroute
       valid_lft 172754sec preferred_lft 172754sec
    inet6 fe80::db63:c3ff:b814:f4e2/64 scope link noprefixroute
       valid_lft forever preferred_lft forever

[fge@fed ~]$ ip -6 route show dev ens10
2001:db8:1::1f proto kernel metric 102 pref medium
2001:db8:1::/64 proto ra metric 102 pref medium
fe80::/64 proto kernel metric 102 pref medium
default via fe80::be2b:a3e2:2b5a:2e7d proto ra metric 102 pref medium

[fge@fed ~]$ cat /etc/resolv.conf |grep :
nameserver fe80::be2b:a3e2:2b5a:2e7d%ens10
nameserver 2001:db8:1::1

[fge@fed ~]$ host -t AAAA fed 2001:db8:1::1
Using domain server:
Name: 2001:db8:1::1
Address: 2001:db8:1::1#53
Aliases:

fed has IPv6 address 2001:db8:1::1f

Server side configuration: radvd

The radvd provide feature-rich support of IPv6-RA and allowing you to do those thins which dnsmasq does not support yet:

  • Advertise more routes.
  • Customize IPv6-RA options.

For more detail, please refer to manpage of dnsmasq.

interface eth1
{
    AdvSendAdvert on;
    MinRtrAdvInterval 30;
    MaxRtrAdvInterval 100;
    prefix 2001:db8:1::/64 {    # Host will use SLAAC for this prefix
        AdvOnLink on;
        AdvAutonomous on;
        AdvRouterAddr off;
    };
    route ::/0 {};              # Host will get default gateway
    route 2001:db8:f::/64 {};   # Host will get this optional stateless route
    RDNSS 2001:db8:1::1 {};     # Host will get nameserver as 2001:db8:1::1
};

Client side configuration using nmstate

The nmstate project provides simple way to set network state.

For example: to enable auto configuration on eth1 for IPv4 and IPv6:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21

echo '
---
interfaces:
- name: eth1
  type: ethernet
  state: up
  ipv4:
    dhcp: true
    auto-dns: true
    auto-gateway: true
    auto-routes: true
    enabled: true
  ipv6:
    autoconf: true
    dhcp: true
    auto-dns: true
    auto-gateway: true
    auto-routes: true
    enabled: true
  mtu: 1500
' | sudo nmstatectl set -

Then you may use command nmstatectl show eth1 to get the current network state of eth1.

Client side configuration using NetworkManager

With NetworkManager daemon running, the below simple command would enable both IPv4 and IPv6 automatic configuration on eth1:

1
2
3
4
5

sudo nmcli connection add type ethernet \
    ifname eth1 \
    +connection.id auto-eth1 \
    +ipv4.method auto \
    +ipv6.method auto

Client side configuration manually

For IPv6/IPv4 DHCP, you need to use dhclient, dhcpcd or NetworkManager or any DHCP client tools.

For IPv6 Router Advertisement, you may enable/disable it via

1
2
3
4

# Disable
sudo sysctl -w net.ipv6.conf.eth1.accept_ra=0
# Enable
sudo sysctl -w net.ipv6.conf.eth1.accept_ra=1

You may also need to enable IPv6 manually

1

sudo sysctl -w net.ipv6.conf.eth1.disable_ipv6=0